Heq.Tech

Inspiration

Despite tougher DUI laws and safety campaigns, drivers remain vulnerable with no clear way of knowing if they’re entering high-risk zones. With over 1.19M deaths globally due to road crashes and 30% involving alcohol-impaired drivers, we wanted to create a solution that proactively warns both the public and authorities—before it’s too late.

What it does

Heq. tech leverages public CCTV traffic feeds to detect reckless or impaired driving behavior using AI. Once flagged, our system:

Notifies police with real-time location data
Maps dangerous driving zones as public heatmaps, for the public
Provides users with alerts and historical risk data for safer route planning

How we built it

We built Heq. tech using:

YOLOv8 for vehicle detection
OpenCV for tracking and frame preprocessing
A custom GRU-based anomaly detector to identify impaired driving patterns
A Flask backend API and MongoDB database for storage and querying
A React Native mobile app to deliver alerts and heatmaps to users

Challenges we ran into

Video quality and variability in public traffic cams made detection inconsistent
Displaying clean and accurate heatmap visualizations while preserving user privacy

Accomplishments that we're proud of

Successfully integrated YOLOv8 + GRU in a real-time anomaly detection pipeline
Designed a clean, functional mobile app with real-time alerting and interactive heatmaps
Developed a working MVP that brings both social awareness and technical complexity to life

What we learned

How to preprocess and stream frame data into sequential models like GRUs
The importance of balancing latency and accuracy in real-time applications
How to design with scalability and ethics in mind, especially around public surveillance